Computational simulation has become a widely accepted research methodology in the social sciences, particularly in the study of organizations. In this paper, we first replicate a computational simulation of exploration and exploitation in organizational learning by James G. March). We then modify and extend the replication with respect to the theory of social systems of Niklas Luhmann. While March's original simulation model features both knowledge and learning, we complement his model with Luhmann's concept of memory, here understood as the continuous discrimination between forgetting and remembering. In this light, it can be shown that forgetting serves as a source of dynamic instability for an organization. It inhibits paralysis caused by inconsistencies in the organization's history and thereby represents an enabling condition for new irritability and thus learning.